Ataxia and Purkinje cell degeneration in mice lackingthe CAMTA1 transcription factorChengzu Longa,1, Chad E. Gruetera,1, Kunhua Songa,1, Song Qina, Xiaoxia Qia, Y. Megan Konga, John M. Sheltonb,James A. Richardsonc, Chun-Li Zhanga, Rhonda Bassel-Dubya, and Eric N. Olsona

Departments of aMolecular Biology, bInternal Medicine, and cPathology, University of Texas Southwestern Medical Center, Dallas, TX 75390

Contributed by Eric N. Olson, June 19, 2014 (sent for review July 1, 2013)

Members of the calmodulin-binding transcription activator (CAMTA)family of proteins function as calcium-sensitive regulators of geneexpression in multicellular organisms ranging from plants to hu-mans. Here, we show that global or nervous system deletion ofCAMTA1 in mice causes severe ataxia with Purkinje cell degenera-tion and cerebellar atrophy, partially resembling the consequencesof haploinsufficiency of the human CAMTA1 locus. Gene-expressionanalysis identified a large collection of neuronal genes that weredysregulated in the brains of CAMTA1-mutant mice, and elucidationof a consensus sequence for binding of CAMTA proteins to DNArevealed the association of CAMTA-binding sites with many ofthese genes. We conclude that CAMTA1 plays an essential role inthe control of Purkinje cell function and survival. CAMTA1-mutantmice provide a model to study the molecular mechanisms of neu-rodegenerative diseases and for screening potential therapeuticinterventions for such disorders.

CAMTA2 | dimerization | palindromic DNA | neural genes

First identified in plants as regulators of stress and cold-sensitive changes in gene expression, calmodulin-binding tran-

scription activator (CAMTA) transcription factors subsequentlyhave been implicated in signal-dependent gene expression inmetazoans ranging from fruit flies to humans (1–7). CAMTAproteins share a series of functional motifs, including a CG-1DNA-binding motif, a TIG domain implicated in DNA bindingand dimerization of transcription factors, calmodulin-binding IQmotifs, and ankyrin repeats that mediate oligomerization (1).There are two mammalian CAMTA genes, CAMTA1 and

CAMTA2, which share extensive homology and partially over-lapping expression patterns. CAMTA1 and 2 are highly expressedin the brain with lower levels of expression in the heart (7). Pre-viously, we reported that CAMTA2 plays an important role inpathological cardiac remodeling, which occurs in response to ab-errant calcium signaling (7). CAMTA2-KO mice are viable and donot show obvious phenotypes but are resistant to cardiac stress (7).The functions of CAMTA1 have not been analyzed in mice.However, haploinsufficiency of CAMTA1, caused by mutationsand genomic rearrangements, has been shown to cause sporadicand familial nonsyndromic intellectual deficiency and childhoodataxia in humans (8, 9). The mechanistic basis of these abnor-malities has not been defined. CAMTA1 also has been implicatedas a tumor suppressor in neuroblastoma and glioblastoma (10, 11).To explore the functions of CAMTA1 in vivo, we generated

mice with a conditional loss-of-function allele. Global deletion ordeletion in the nervous system of CAMTA1 resulted in severeataxia, accompanied by Purkinje cell degeneration, decreasedcerebellar size, and disruption of neuronal gene expression.Through an unbiased determination of the optimal CAMTADNA-binding site, we found CAMTA-binding sites to be asso-ciated with a high percentage of dysregulated neuronal genes inthe cerebellum of CAMTA1-mutant mice. We conclude thatCAMTA1 is an essential regulator of Purkinje cell function andsurvival. CAMTA1-mutant mice provide a model for understandingthe mechanistic basis of ataxias and cerebellar function.

ResultsGeneration of CAMTA1-KO Mice. The mouse CAMTA1 gene(OTTMUSG00000010309 in Vega Genome Browser), locatedon chromosome 4, contains at least 20 exons with multiple splicevariants, spanning more than 800 kb of DNA. To investigatethe functions of CAMTA1 in vivo, we generated a conditionalmutant allele of the gene by inserting a loxP site in the intronupstream of exon 9 and a FLP recognition target (FRT)-neo-FRT-LoxP cassette immediately downstream of exon 9 throughhomologous recombination in ES cells (Fig. S1A). Cre-mediatedrecombination of this allele introduced a frame-shift mutationafter amino acid 284 of CAMTA1, eliminating the majority ofthe functional domains and creating a loss-of-function allele.Based on mutagenesis studies of conserved domains in CAMTA2(7), the CG-1 domain contained within the residual portion ofthe mutated CAMTA1 gene would not be expected to functionas a dominant-negative mutant. Targeting of CAMTA1 in EScells and mice was confirmed by Southern blot analysis and PCR(Fig. S1 B and C).Male mice heterozygous for the conditional CAMTA1fl/+ allele

were bred to females carrying a CAG-Cre transgene, whichmediates gene excision at the zygote stage (12), generatingCAMTA1+/− offspring. Mice heterozygous for the CAMTA1+/−

allele appeared phenotypically normal, although we did notsubject these mice to detailed phenotypic analysis. CAMTA1+/−

mice were intercrossed to generate CAMTA1−/− (CAMTA1-KO)mice. Homozygosity of the CAMTA1-KO allele resulted in post-natal death before age 4 wk (Table S1). CAMTA1-KO mice wereseverely runted and displayed ataxia (Fig. S2A and Movie S1).Although CAMTA1 has been suggested to function as a tumor

Significance

Neurodegenerative diseases are debilitating conditions thatresult from degeneration of the nervous system causing symp-toms including ataxia and/or dementia. Calmodulin-bindingtranscription activator 1 (CAMTA1) is a transcription factor en-riched in the brain with the highest levels of expression in thecerebellar granular layer and Purkinje cells, midbrain, pons, andhippocampus. When CAMTA1 is deleted from the nervous sys-tem of mutant mice, we observe a clear loss of Purkinje cells andreduced cerebellar size. Thus, we conclude that CAMTA1 playsa predominant role in the maturation and survival of cerebellarneurons rather than in the initial development of these cells. Theneurological abnormalities associated with CAMTA1-mutantmice allow mechanistic analysis of these abnormalities anddisease modeling.

Author contributions: C.L., C.E.G., K.S., R.B.-D., and E.N.O. designed research; C.L., C.E.G.,K.S., S.Q., X.Q., and J.M.S. performed research; C.L., C.E.G., K.S., Y.M.K., J.M.S., J.A.R., C.-L.Z.,R.B.-D., and E.N.O. analyzed data; and C.L., C.E.G., K.S., R.B.-D., and E.N.O. wrotethe paper.

The authors declare no conflict of interest.1C.L., C.E.G., and K.S. contributed equally to this work.2To whom correspondence should be addressed. Email: Eric.Olson@utsouthwestern.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1411251111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1411251111 PNAS | August 5, 2014 | vol. 111 | no. 31 | 11521–11526

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Nov

embe

r 21

, 202

0

suppressor within neurons and glial cells (10, 11), we never ob-served any tumors in CAMTA1-KO mice before death.

Neuronal Expression of CAMTA1. Previous studies reported strongexpression of CAMTA1 in the cerebellum, as well as in the hip-pocampi and olfactory bulbs of embryonic and neonatal mice (9).Consistent with these findings, in situ hybridization of adult brainsections with CAMTA1-specific probes showed that CAMTA1transcripts were highly enriched in the cerebellum (Fig. 1A).Within the cerebellum, CAMTA1 expression was enriched inPurkinje cells, and the granule layer and was absent in KO mice(Fig. 1B).

Motor Abnormalities of CAMTA1fl/fl, Nestin-Cre Mice. To determinethe cellular basis of the neurological phenotype of mice withglobal deletion of CAMTA1, we generated CAMTA1 deletions bycrossing CAMTA1fl/flmice with mice bearing aNestin-Cre transgene,which is expressed throughout the nervous system (13). RT-PCRanalysis of total RNA from cerebella of CAMTA1fl/fl, Nestin-Cremice (referred to as “CAMTA1-nKO mice”) and in situ hybridiza-tion of CAMTA1-nKO brain sections revealed that CAMTA1 wasdeleted successfully (Fig. 1B and Fig. S1D).Consistent with the ataxia observed in mice with global de-

letion of CAMTA1, specific deletion of CAMTA1 in the nervoussystem with Nestin-Cre resulted in severe ataxia as assessed bymultiple motor tasks and general observation (Fig. 2 and Fig. S2B–D). The ataxic phenotype was observed in every CAMTA1-nKO mouse over the age of 3 mo. Nestin-Cre transgenic micecontaining WT CAMTA1 alleles did not display ataxia (13).CAMTA1-nKO mice were able to breed at young ages but did soat a reduced frequency and did not show reduced survival rel-ative to CAMTA1fl/fl (control) littermates up to age 1 y. Thesevere runting observed upon global deletion of CAMTA1 wasless pronounced and was seen in only a subset of nKO mice,

suggesting that CAMTA1 has additional functions beyond thecell types in which Nestin-Cre is active.To assess motor coordination quantitatively, nKO mice and

CAMTA1fl/fl littermates were subjected to rotarod testing at 6and 12 wk of age. In multiple trials, CAMTA1-nKO mice dis-played a significant reduction in time spent on the rotarod (Fig. 2A and B). By age 6 wk CAMTA1-nKO mice also experiencedhighly significant delays in the time needed to cross balancebeams of multiple sizes on several successive attempts (Fig. 2Cand Fig. S2 B and C). The loss of motor coordination was evidentin all CAMTA1-nKO mice and increased in a rostro-caudalgradient as depicted in Movie S2.We also deleted the conditional CAMTA1 allele using a Pax3-

Cre knockin allele, which is expressed in embryonic somites andthe dorsal neural tube, where precursors of the central nervoussystem originate (14). These mice also developed severe ataxiaby 6 wk of age. Therefore we focused our subsequent studies onthe nKO mice.

Progressive Cerebellar Atrophy and Degeneration of Purkinje Cells inCAMTA1-nKO Mice. To delineate the cellular basis of the ataxicphenotype of CAMTA1-nKO mice, adult brains from CAMTA1fl/fl

and CAMTA1-nKO mice were analyzed. The behavioral results ofthe accelerating rotarod tests strongly suggested impairments inmotor coordination; however, the severity of the phenotype maymask striatal/dopamine-dependent motor skill learning. No grosschanges in neural architecture were observed in the forebrain ormidbrain of nKO mice (Fig. 3A). Therefore, we focused our at-tention on cerebellar structure. Deletion of CAMTA1 did not alterthe normal cerebellar architecture at 2–3 wk of age but caused adramatic loss of Purkinje cells and a decrease in cerebellar size atage 3 mo, concomitant with the loss of motor coordination (Fig. 3A and B). The loss of Purkinje cells, readily visualized by immu-nostaining for Calbindin (Fig. 3 C and D), also correlated witha significant decrease in the width of the molecular and granularlayers within the cerebellum of CAMTA1-nKO mice (Fig. 3 E–H).

Identification of a DNA Recognition Motif for CAMTA Proteins.CAMTA proteins from fruit flies and plants bind to a CG-richDNA sequence via the CG-1 motif near their N termini (4, 6). Toinvestigate the DNA-binding properties of mammalian CAMTAproteins, we fused various regions of mouse CAMTA1 andCAMTA2 to the maltose-binding protein (MBP). Among variousCAMTA–MBP fusions, we were able to obtain stable protein onlyby using a cDNA fragment coding for the first 630 amino acids ofmouse CAMTA2, which encompasses the CG motif, transcriptionactivation domain, and TIG motif (Fig. S3). The CG-1 and TIGmotifs of CAMTA1 and 2 are highly homologous (Fig. S3).The bacterially expressed MBP–CAMTA2 fusion protein

was used in iterative gel mobility shift assays with random oli-gonucleotide pools. Following four rounds of selection, weidentified a consensus sequence for CAMTA-DNA binding ofPyGCANTGCG (Fig. 4 A and B and Fig. S4A). Here “Py”represents T or C, and “N” represents A, T, or G. Gel mobilityshift assays showed that the mobility of an oligonucleotide probecontaining the CAMTA consensus sequence was retarded byMBP–CAMTA2, but not by MBP alone (Fig. S4B). Mutation ofthe core sequence CGCATTGCG to CGgtTcaaG completelyabolished the interaction of MBP–CAMTA2 and DNA, whereas amutation in the sequence adjacent to the core consensus didnot abolish the interaction between MBP–CAMTA2 and DNA(Fig. S4B). However, the latter mutation reduced DNA bindingby MBP–CAMTA2, indicating that the adjacent sequences in-fluence interaction with CAMTA protein.To characterize the consensus DNA-binding sequence of

CAMTA proteins further, we generated 15 mutations of the coreconsensus sequence (Fig. 4C). Gel-shift assays using full-lengthCAMTA2 translated in vitro showed that four mutations—M1,M6, M11, and M12—did not abolish the interaction of CAMTAproteins with the DNA, whereas three mutations—M2, M8, andM10—greatly decreased the interaction (Fig. 4 C and D). All

Fig. 1. Expression of CAMTA1 in mouse cerebellum. In situ hybridizationanalysis of sagittal section of adult mouse brain (2 mo old) of (A) WT and (B)CAMTA1-nKO shows expression pattern of CAMTA1 in the cerebellum. Theboxed region in the upper panels is enlarged in the lower panels. Signalshows expression of CAMTA1 in the WT Purkinje cells. Minimal expression ofCAMTA1 is seen in the Purkinje cell layer of CAMTA1-nKO mice. Red arrowsindicate the localization of Purkinje cells. (Scale bars: Upper, 600 μm; Lower,100 μm.)

11522 | www.pnas.org/cgi/doi/10.1073/pnas.1411251111 Long et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 21

, 202

0

other mutations (M3–M5, M7, M9, and M13-14) abolish in-teraction of CAMTA proteins with DNA (Fig. 4 C and D). Theseresults suggested that the consensus DNA-binding sequence ofCAMTA could be (T/C)GCANTGCG, where “N” stands forany nucleotide.Adjacent to the consensus DNA-binding sequence of CAMTA,

we noticed A/T-rich sequences that are known to increase DNAflexibility and to be crucial for target site recognition of GC box-binding proteins (15). To test the relevance of the A/T-richsequences in the CAMTA-binding site, we generated nine muta-tions (f1 to f9) with reduced length of A/T-rich sequence (Fig.4E). Four mutations, f6–f9, greatly decreased the interaction ofCAMTA protein and the DNA probe (Fig. 4F), suggesting thatthe adjacent A/T-rich sequences also are crucial for the coresequence binding by CAMTA.To examine the ability of CAMTA proteins to activate tran-

scription through the identified consensus sequence, we con-structed a reporter (1×CAMTA-luc), consisting of one copy ofthe CAMTA consensus sequence CGCATTGCG linked to aTATA box of the E1b virus promoter. In transfected COS cells,both CAMTA1 and CAMTA2 were able to activate reporter

gene expression, with CAMTA1 conferring higher activationthan CAMTA2 (Fig. S4C). Changing the core CGCATTGCGto CGgtTcaaG abolished the transcriptional activity of CAMTAproteins on the reporter (Fig. S4C), indicating that the consensusbinding site CGCATTGCG is required for transcriptional acti-vation by CAMTA proteins.Replacement of the conserved lysine K108 in the CG-1 domain

of CAMTA2 with alanine or of the corresponding lysine K141 inCAMTA1 with glutamic acid completely disrupted the DNA-binding activity of CAMTA proteins (Fig. S4 D–F). These mu-tant proteins also were unable to activate the 1×CAMTA-lucreporter in COS cells (Fig. S4G).

Dimerization of CAMTA on the Palindromic Core Sequence. TheCAMTA consensus binding site shows a symmetry surroundingthe central variable N position, suggesting a bipartite mode ofDNA binding (Fig. 5A). Gel mobility shift assays were performedwith FLAG-tagged N terminus (amino acids 1–625) of CAMTA2(termed MBP-FLAG-CAMTA) incubated with DNA probe con-taining the CAMTA-binding site to generate two bands on thegel (Fig. 5B). Both bands showed a supershift when incubatedwith anti-FLAG antibody. The ratio of the intensity of the upperband to that of the lower band was decreased when less CAMTAprotein was added to the same amount of DNA probe. Only thelower band was detectable when the least amount of CAMTAprotein was added. These results suggest that the higher band is acomplex of a CAMTA homodimer and the DNA probe, whereasthe lower band is a complex of a CAMTA monomer and theDNA probe.To confirm dimerization of CAMTA on DNA, we performed

an in vitro pull-down assay using in vitro-translated FLAG-CAMTA as pull-down “bait” and in vitro-translated 35S-labeledmyc-CAMTA as “prey” to detect interactions. Equal amounts ofFLAG-CAMTA and 35S-myc-CAMTA were mixed with DNA andanti-FLAG beads (Fig. 5C). If CAMTA forms a dimer when bindingto DNA, 35S-myc-CAMTA should be pulled down with anti-FLAG beads (Fig. 5C, Left). If CAMTA forms a monomer whenbinding DNA, anti-FLAG beads cannot capture 35S-myc-CAMTA(Fig. 5C, Right). The results show that 35S-myc-CAMTA can bepulled down with anti-FLAG beads, but not with IgG beads, in

Fig. 2. Motor deficits of CAMTA1-nKO mice. (A and B) Motor coordinationperformance on a rotarod with slow acceleration from 5–25 rpm over 5 minwas assessed in CAMTA1fl/fl control and CAMTA1-nKO mice at age 6 (A) and12 (B) wk. n = 6 control mice, n = 5 CAMTA1-nKO mice; *P < 0.05. Resultsfrom four independent trials are shown. (C) CAMTA1-nKO and CAMTA1fl/fl

control mice at age 6 wk were evaluated for motor deficits using the beamwalk test. The time needed to traverse an 18-mm round wooden rod wasrecorded in multiple trials. n = 6 control mice, n = 4 CAMTA1-nKO mice; *P <0.001. Results from three independent trials are shown.

Fig. 3. Degeneration of Purkinje cells and cerebellar atrophy in CAMTA1-nKO mice. (A) Gross morphology of brains from adult WT and CAMTA1-nKO mice.(B) H&E stain of sagittal sections of brain from 3-wk-old (Upper) and adult (Lower) WT and CAMTA1-nKO mice. (Scale bar, 400 μm.) (C) Calbindin-D28Kimmunohistochemical staining (red) of Purkinje cells in the cerebellum of 6-mo-old CAMTA1-nKO and littermate control mice. Purkinje cells are uniformlyorganized in the cerebellum of the control mouse but are reduced and disarrayed in the cerebellum of the CAMTA1-nKO mouse. Nuclei are stained with DAPI(blue). gl, granular layer. (Scale bar, 100 μm.) (D) Quantification of the size of calbindin-positive Purkinje cells in the cerebellum (n = 3). (E and F) H&E stainingof sagittal sections of the brain from CAMTA1-nKO and littermate control mice at age 2–3 wk (E) and 3 mo (F). White arrows point to Purkinje cells. (Scale bar,400 μm Upper, 40 μm Lower.) n = 6; *P < 0.05. gl, granular layer; ml, molecular layer. (G and H) Quantification of the Purkinje cells of CAMTA1-nKO andlittermate control mice at age 2–3 wk (G) and 3 mo (H). Purkinje cell number is not reduced in CAMTA1-nKO mice at age 2–3 wk but is decreased at age 3 moalong with reductions in the width of the molecular layer and granular layer. Data are presented as mean ± SD.

Long et al. PNAS | August 5, 2014 | vol. 111 | no. 31 | 11523

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Nov

embe

r 21

, 202

0

the presence of FLAG-CAMTA. In the absence of FLAG-CAMTA, 35S-myc CAMTA was not pulled down by anti-FLAG beads. Without the DNA probe, the interaction be-tween 35S-myc CAMTA and FLAG-CAMTA was decreasedgreatly (Fig. 5D). Taken together, these data support theconclusion that CAMTA binds to its palindromic DNA coresequence as a dimer.

Changes in Neuronal Gene Expression in CAMTA1-nKO Mice. To beginto define the molecular basis of CAMTA1 function in the ner-vous system, we compared the gene-expression profiles of cere-bella of control and nKO mice by microarray analysis. Thesestudies identified 203 genes that were dysregulated by at least 1.5-fold in CAMTA1-nKO cerebella (Fig. 6). Of these 203 genes, 69were down-regulated and 134 were up-regulated in CAMTA1-nKO mice. Gene Ontology analysis revealed that 84 of the 203genes were involved in neuronal functions or in the protection ofneurons from apoptosis (Fig. 6). Selected genes containingputative CAMTA consensus sites within 5 kb upstream of thetranscriptional start site that are down- and up-regulated inCAMTA1-nKO cerebellum are listed in Table S2.Analysis of DNA sequences associated with genes down-

regulated in the CAMTA1-nKO cerebella revealed that atleast 11 genes are associated with CAMTA consensus sites(T/C)GCANTGCG (Table S2). Interestingly, Snhg11 (NM_175692.3Riken cDNA A930034L06), the most down-regulated gene inCAMTA1-nKO cerebella, is highly expressed in Purkinje cellswithin the cerebellar granular layer, corresponding to the expres-sion pattern of CAMTA1. Another gene strongly down-regulatedin nKO cerebella, Gtl2, has been shown to function as a tumorsuppressor (16). Numerous other CAMTA1-sensitive genes thatwe found to be associated with consensus CAMTA-binding sitesare involved in stress responsiveness, signal transduction, and

protein metabolism. Among genes up-regulated in CAMTA1-nKOcerebellum, transthyretin (TTR) was up-regulated 51-fold in nKOmice (Table S2). TTR has been implicated in neuroprotection (17,18). The CAMTA consensus site does not exist within 5 kb up-stream of the TTR gene; therefore up-regulation of TTR may bea secondary response to neurodegeneration in CAMTA1-nKObrain. Somatostatin (Sst), a multifunctional peptide, also was up-regulated in nKO brain. Increased expression of Sst in neurons isassociated with motor coordination disabilities in Huntington’schorea (19). Sst accumulation increases in response to β-amyloid–induced neurotoxicity in cortical neurons (20). Up-regulation ofgenes in response to neuronal cell death partially suggestsneurodegeneration in CAMTA1-nKO mice. Thus, deletion ofCAMTA1 in the nervous system dysregulates the expressionof a broad collection of genes required for survival and ho-meostasis of neuronal cells, including Purkinje cells and gran-ule neurons.

DiscussionThe results of this study demonstrate an essential role of theCAMTA1 transcription factor in maintaining cerebellar functionin mice and suggest that this calcium-sensitive transcriptionfactor is required for coordination of gene expression and sur-vival of Purkinje cells. The absence of CAMTA1 in mice resultsin severe ataxia and neuronal atrophy. Through the identifica-tion of the consensus DNA-binding site for mammalian CAMTAproteins and the analysis of dysregulated cerebellar genes inCAMTA1-KO mice, we identified a large collection of genes thatlikely mediate the essential functions of CAMTA1 in cerebellardevelopment and function. Given the many genes that are ap-parently regulated by CAMTA1 in the cerebellum, it seems likelythat the ataxic abnormalities of CAMTA1-nKO mice reflect the

Fig. 4. Identification of the CAMTA DNA-binding site and adjacent A/T-rich sequence. (A) Strategy for selecting CAMTA-binding DNA sequences using a DNAoligonucleotide library containing random nucleotides at 18 variable positions flanked by primers A and B. 32P-labeled double-stranded oligo-pools weregenerated by primers A and B and incubated with MBP or MBP-FLAG-CAMTA2 (amino acids 1–625). The DNA–CAMTA complex was resolved on a nativepolyacrylamide gel (Gel shift). DNA was recovered from the gel and labeled for the next round of selection. (B) Gel mobility shift of the DNA–CAMTA complexon a 4% native polyacrylamide gel. DNA recovered from the gel was used for the next round of selection. Four cycles of selection were performed. C1, the firstcycle of selection; C2, the second cycle of selection; C4, the fourth cycle of selection. Anti-FLAG was used for supershift. (C) Sequences of the CAMTA consensussite (WT) and mutations (M1–M15) used for gel mobility shift assays. The relative binding of CAMTA to the different oligos is summarized. (D) Gel mobilityshift assays to detect interactions between CAMTA2 and the oligos shown in C. (E) CAMTA consensus site with the adjacent A/T-rich sequence mutations(green) used for gel mobility shift assays. The relative binding of CAMTA to the different oligos is summarized. (F) Gel mobility shift assays to detectinteractions between CAMTA2 and the oligos listed in E. Full-length CAMTA2 was translated in vitro using pcDNA (Life Technologies)-CAMTA2 as template.Anti-FLAG antibody was used for supershift. pcDNA empty vector was used as template in lysate controls.

11524 | www.pnas.org/cgi/doi/10.1073/pnas.1411251111 Long et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 21

, 202

0

combined functions of many of these genes rather than that ofa single downstream target gene.CAMTA1 is highly enriched in the brain, with the highest

levels of expression in the cerebellar granular layer and Purkinjecells, midbrain, pons, and hippocampus. During the early stagesof postnatal development, the cerebellum undergoes a dramaticincrease in size and developmental changes that culminate inmaturation of the adult cerebellum (21). The morphologicalchanges occurring in the cerebellum include maturation ofPurkinje cells with an increase in dendritic arborization, con-comitant with an increase in the molecular layer and granularlayer within the cerebellum. Based on histological analysis, theanatomy and size of the cerebellum appeared normal in 2- to3-wk-old CAMTA1-nKO mice. However, we observed a clearloss of Purkinje cells and reduced cerebellar size in mutant mice byage 3 mo. Thus, we conclude that CAMTA1 plays a predominantrole in the maturation and survival of cerebellar neurons ratherthan in the initial development of these cells.CAMTA1 is one of many genes located within the 1p36

chromosomal locus, which corresponds to a common chromo-somal breakpoint associated with a broad range of clinical fea-tures, including neurological defects, in humans (22). In thisregard, multiple SNPs that correlate to cognitive abnormalitieshave been identified in the CAMTA1 gene (8). The importanceof CAMTA1 for normal neurological function was highlightedfurther by the identification of human subjects with heterozygouschromosomal rearrangements in the CAMTA1 locus who presentwith congenital cerebellar ataxia, in accord with our findings inmice (9). It should be emphasized, however, that these humanphenotypes result from heterozygous CAMTA1 mutations, whereasthe abnormalities observed in mice result from homozygous genedeletion. There also are some distinctions between the neuro-logic phenotypes seen in our CAMTA1-nKO mice and humans

with CAMTA1 haploinsufficiency. In particular, the ataxia seenin humans does not appear to be progressive and has not beenshown to be accompanied by neuronal degeneration as seen inour homozygous mutant mice.Although CAMTA1 is expressed predominantly in the

brain, it also is expressed in the heart (7), and it has beenreported that CAMTA1 is required for differentiation of adultstem cells into myocardial cells (23). We did not detect any overtcardiac abnormalities in mice with global deletion of CAMTA1.Interestingly, however, a polymorphism in CAMTA1 revealedan association in patients with cardiovascular disease who hadpoor cognitive performance, suggesting the potential involve-ment of CAMTA1 in both cardiovascular and neural func-tion (24).Our previous studies demonstrated regulation of CAMTA

proteins by association with class II histone deacetylases (HDACs)(7). In this regard, HDAC4 associates with ataxin-1 and myocyteenhancer transcription factor 2 (MEF2) to regulate neuronalsurvival (25). The MEF2/HDAC5/calcium/calmodulin-dependentprotein kinase II (CaMKII)-signaling pathway also regulatessurvival of cerebellar granule cells (26), and our results show thatdeletion of CAMTA1 leads to a decrease in the granule cell layerin adult mice. Thus, we speculate that the influence of CAMTAon neuronal survival is subject to signal-dependent regulation bythe MEF2/HDAC5/CaMKII pathway.In plants, CAMTA proteins activate stress-responsive genes by

binding to a CG-rich motif, CG(C/T)G (2, 3, 5, 6). We found thatmammalian CAMTA proteins also bind to a CG-rich motif,(T/C)GCANTGCG, although this sequence is distinct from thatof plant CAMTA binding. In the absence of CAMTA1, nu-merous genes are dysregulated in the cerebellum. Among thesegenes, Eif4a2, Fbxl3a, and Magee1 are involved in protein syn-thesis and degradation. Neurodegenerative disorders often areassociated with the accumulation of misfolded disease-specificproteins, causing repression of protein translation (27, 28),which has been associated with synaptic failure and neuronalloss (27). It will be interesting to investigate whether loss ofprotein quality is a major trigger to degeneration of Purkinjecells and other neurons in the absence of CAMTA1. Thefunctions of numerous other CAMTA1-dependent genes, suchas Snhg11, Snurf, and Snrpn, are currently unknown. How thesegenes affect the functions of cerebellar neurons and other cellsof the nervous system in which CAMTA1 is expressed remainsto be determined.The neurological abnormalities associated with CAMTA1-

mutant mice allow mechanistic analysis of these abnormalitiesand disease modeling. Further studies of the CAMTA-sig-naling pathway in the nervous system may contribute to theidentification of therapeutic targets for intervention in neu-rodegenerative diseases. Understanding how the derange-ments in neuronal gene expression associated with CAMTA1

Fig. 5. Dimerization of CAMTA on the palindromic core sequence. (A)Schematic of CAMTA dimers on the palindromic core sequence. (B) Gelmobility shift assay consisting of recombinant protein of the N terminus(amino acids 1–625) of CAMTA (MBP-FLAG-CAMTA) incubated with its DNA-binding site generates two bands on the gel. Both bands were supershiftedby anti-FLAG antibody. When decreasing amounts of CAMTA protein areadded (designated by the black triangle), the ratio of the intensity of theupper band to the intensity of the lower band is decreased. The higher bandis a complex of a CAMTA homodimer and DNA probe. The lower band isa complex of a CAMTA monomer and DNA probe. (C) Schematic of the invitro pull-down assay of dimerization of CAMTA. (D) Pull-down assay show-ing 35S-labeled, myc-tagged CAMTA incubated with FLAG-CAMTA and theDNA core sequence could be pulled down with anti-FLAG beads, whereas35S-labeled, myc-tagged CAMTA alone or with DNA was not pulled down byanti-FLAG beads.

Fig. 6. Identification of dysregulated genes in CAMTA1-nKO cerebellum.Microarray analysis of RNA isolated from cerebella of control and CAMTA1-nKO mice identified 203 genes that were up- or down-regulated by at least1.5-fold in CAMTA1-nKO mice.

Long et al. PNAS | August 5, 2014 | vol. 111 | no. 31 | 11525

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Nov

embe

r 21

, 202

0

deletion relate to those associated with other transcription factormutations, such as those in TATA box-binding protein (29, 30),also will be of interest.

Materials and MethodsGeneration of CAMTA1 Conditional-KO Mice. CAMTA1 conditional-KO micewere generated by inserting sites for Cre-mediated recombination upstreamand downstream of exon 9 through homologous recombination, as de-scribed in SI Materials and Methods. Southern blot analysis was performedas previously described (7). All animal procedures were approved by theUniversity of Texas Southwestern Medical Center Institutional Animal Careand Use Committee.

Histological and Immunohistochemical Analysis. Brains were harvested fromCAMTA1 control and nKO mice. Histology and immunohistochemistry wereperformed as described previously (31). Frozen sections were stained withanti–calbindin-D28K antibody (1:500) (C9848; Sigma). The size of calbindin-positive Purkinje cells was assessed using Stereo Investigator (MBF Bioscience).

In Situ Hybridization. In situ hybridization was performed as described (32).Briefly, an 807-bp cDNA fragment located in exon 9 of the mouse CAMTA1gene was isolated by PCR and subcloned into a pGEM-T Easy vector (forwardprimer: TGTCCGAGGTCACTAACGAG; reverse primer: GAGGTGCAAGGAG-GAAGTAGA). Digoxin-labeled sense or antisense riboprobes were generatedby in vitro transcription with SP6 or T7 RNA polymerase (Roche). In situhybridization then was performed on sagittal cryostat sections of 16 μm.

Beam Walk and Rotarod Performance Test. Neurobehavioral tests were per-formed at the University of Texas Southwestern Rodent Behavior Core Fa-cility, as described in SI Materials and Methods.

CAMTA Site Selection and Transfection Assays. DNA-binding site selectionusing CAMTA2-MBP fusion proteins and transfection assays were performedas described in SI Materials and Methods.

RNA and Microarray Analysis. RNAwas isolated frommouse brain using TRIzolreagent and was used for microarray analysis performed by the University ofTexas Southwestern Microarray Core Facility using the MouseWG-6 v2.0BeadChips (Illumina) or was used for quantitative PCR as described in SIMaterials and Methods.

Reporter Activity Assays. Plasmid constructs and transfection assays for lu-ciferase reporter assays are described in SI Materials and Methods.

Statistics. Values are presented as the mean ± SE. Student t test was per-formed for paired analysis. Repeated-measure two-way ANOVA with Tukey’smultiple comparisons post hoc test was used to analyze behavior studies.P < 0.05 was considered statistically significant.

ACKNOWLEDGMENTS. We thank Jose Cabrera for graphics and Shari Birn-baum at the University of Texas Southwestern Rodent Behavior Core Facilityfor mouse behavior analysis. This work was supported in part by grants fromNational Institutes of Health and by Grant I-0025 from the Robert A. WelchFoundation (to E.N.O.). C.E.G. was supported by Fellowship 7-09-CVD-04from the American Diabetes Association.

1. Bouché N, Scharlat A, Snedden W, Bouchez D, Fromm H (2002) A novel family ofcalmodulin-binding transcription activators in multicellular organisms. J Biol Chem277(24):21851–21861.

2. Doherty CJ, Van Buskirk HA, Myers SJ, Thomashow MF (2009) Roles for ArabidopsisCAMTA transcription factors in cold-regulated gene expression and freezing tolerance.Plant Cell 21(3):972–984.

3. Galon Y, et al. (2008) Calmodulin-binding transcription activator (CAMTA) 3 mediatesbiotic defense responses in Arabidopsis. FEBS Lett 582(6):943–948.

4. Han J, et al. (2006) The fly CAMTA transcription factor potentiates deactivation ofrhodopsin, a G protein-coupled light receptor. Cell 127(4):847–858.

5. Kim Y, Park S, Gilmour SJ, Thomashow MF (2013) Roles of CAMTA transcription fac-tors and salicylic acid in configuring the low-temperature transcriptome and freezingtolerance of Arabidopsis. Plant J 75(3):364–376.

6. Pandey N, et al. (2013) CAMTA 1 regulates drought responses in Arabidopsis thaliana.BMC Genomics 14:216.

7. Song K, et al. (2006) The transcriptional coactivator CAMTA2 stimulates cardiacgrowth by opposing class II histone deacetylases. Cell 125(3):453–466.

8. Huentelman MJ, et al. (2007) Calmodulin-binding transcription activator 1 (CAMTA1)alleles predispose human episodic memory performance. Hum Mol Genet 16(12):1469–1477.

9. Thevenon J, et al. (2012) Intragenic CAMTA1 rearrangements cause non-progressivecongenital ataxia with or without intellectual disability. J Med Genet 49(6):400–408.

10. Henrich KO, et al. (2011) CAMTA1, a 1p36 tumor suppressor candidate, inhibitsgrowth and activates differentiation programs in neuroblastoma cells. Cancer Res71(8):3142–3151.

11. Schraivogel D, et al. (2011) CAMTA1 is a novel tumour suppressor regulated by miR-9/9*in glioblastoma stem cells. EMBO J 30(20):4309–4322.

12. Sakai K, Miyazaki Ji (1997) A transgenic mouse line that retains Cre recombinaseactivity in mature oocytes irrespective of the cre transgene transmission. BiochemBiophys Res Commun 237(2):318–324.

13. Tronche F, et al. (1999) Disruption of the glucocorticoid receptor gene in the nervoussystem results in reduced anxiety. Nat Genet 23(1):99–103.

14. Engleka KA, et al. (2005) Insertion of Cre into the Pax3 locus creates a new allele ofSplotch and identifies unexpected Pax3 derivatives. Dev Biol 280(2):396–406.

15. Lundin M, Nehlin JO, Ronne H (1994) Importance of a flanking AT-rich region intarget site recognition by the GC box-binding zinc finger protein MIG1. Mol Cell Biol14(3):1979–1985.

16. Zhou Y, Zhang X, Klibanski A (2012) MEG3 noncoding RNA: A tumor suppressor. J MolEndocrinol 48(3):R45–R53.

17. Fleming CE, Saraiva MJ, Sousa MM (2007) Transthyretin enhances nerve regeneration.J Neurochem 103(2):831–839.

18. Stein TD, Johnson JA (2002) Lack of neurodegeneration in transgenic mice over-expressing mutant amyloid precursor protein is associated with increased levelsof transthyretin and the activation of cell survival pathways. J Neurosci 22(17):7380–7388.

19. Beal MF, Uhl G, Mazurek MF, Kowall N, Martin JB (1986) Somatostatin: Alterations inthe central nervous system in neurological diseases. Res Publ Assoc Res Nerv Ment Dis64:215–257.

20. Geci C, How J, Alturaihi H, Kumar U (2007) Beta-amyloid increases somatostatin ex-pression in cultured cortical neurons. J Neurochem 101(3):664–673.

21. Goldowitz D, Hamre K (1998) The cells and molecules that make a cerebellum. TrendsNeurosci 21(9):375–382.

22. Battaglia A (2005) Del 1p36 syndrome: A newly emerging clinical entity. Brain Dev27(5):358–361.

23. Muller-Borer B, et al. (2012) Calcium dependent CAMTA1 in adult stem cell com-mitment to a myocardial lineage. PLoS ONE 7(6):e38454.

24. Miller LA, et al. (2011) CAMTA1 T polymorphism is associated with neuropsychologicaltest performance in older adults with cardiovascular disease. Psychogeriatrics 11(3):135–140.

25. Bolger TA, Zhao X, Cohen TJ, Tsai CC, Yao TP (2007) The neurodegenerative diseaseprotein ataxin-1 antagonizes the neuronal survival function of myocyte enhancerfactor-2. J Biol Chem 282(40):29186–29192.

26. Heidenreich KA, Linseman DA (2004) Myocyte enhancer factor-2 transcription factorsin neuronal differentiation and survival. Mol Neurobiol 29(2):155–166.

27. Moreno JA, et al. (2012) Sustained translational repression by eIF2α-P mediates prionneurodegeneration. Nature 485(7399):507–511.

28. Rubinsztein DC (2006) The roles of intracellular protein-degradation pathways inneurodegeneration. Nature 443(7113):780–786.

29. van Roon-Mom WM, Reid SJ, Faull RL, Snell RG (2005) TATA-binding protein inneurodegenerative disease. Neuroscience 133(4):863–872.

30. Nakamura K, et al. (2001) SCA17, a novel autosomal dominant cerebellar ataxiacaused by an expanded polyglutamine in TATA-binding protein. Hum Mol Genet10(14):1441–1448.

31. Montgomery RL, Hsieh J, Barbosa AC, Richardson JA, Olson EN (2009) Histone de-acetylases 1 and 2 control the progression of neural precursors to neurons duringbrain development. Proc Natl Acad Sci USA 106(19):7876–7881.

32. Qin S, Liu M, Niu W, Zhang CL (2011) Dysregulation of Kruppel-like factor 4 duringbrain development leads to hydrocephalus in mice. Proc Natl Acad Sci USA 108(52):21117–21121.

11526 | www.pnas.org/cgi/doi/10.1073/pnas.1411251111 Long et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 21

, 202

0